Specific Ion Binding to Macromolecules: Effects of Hydrophobicity and

Feb 23, 2008 - Halogen-Ionic Bridges: Do They Exist in the Biomolecular World? Peng Zhou , Yanrong Ren , Feifei Tian , Jianwei Zou and Zhicai Shang. J...
1 downloads 9 Views 175KB Size
Langmuir 2008, 24, 3387-3391

3387

Specific Ion Binding to Macromolecules: Effects of Hydrophobicity and Ion Pairing Mikael Lund,* Robert Va´cha, and Pavel Jungwirth Institute of Organic Chemistry and Biochemistry, Academy of Sciences of the Czech Republic, and Center for Biomolecules and Complex Molecular Systems, FlamingoVo na´ m. 2, CZ-16610 Prague 6, Czech Republic ReceiVed NoVember 1, 2007. In Final Form: December 21, 2007 Using molecular dynamics simulations in an explicit aqueous solvent, we examine the binding of fluoride versus iodide to a spherical macromolecule with both hydrophobic and positively charged patches. Rationalizing our observations, we divide the ion association interaction into two mechanisms: (1) poorly solvated iodide ions are attracted to hydrophobic surface patches, while (2) the strongly solvated fluoride and to a minor extent also iodide bind via cation-anion interactions. Quantitatively, the binding affinities vary significantly with the accessibility of the charged groups as well as the surface potential; therefore, we expect the ion-macromolecule association to be modulated by the local surface characteristics of the (bio-)macromolecule. The observed cation-anion pairing preference is in excellent agreement with experimental data.

Introduction The physicochemical properties of a range of biological and colloidal systems such as macromolecular association, protein activity, denaturation, and so forth1-4 are affected by the presence of salt ions in the surrounding aqueous solution. This influence depends not only on the salt concentration and valency but also on more subtle characteristics of the solvated ions, commonly bundled under empirical terms such as “Hofmeister effects” and “electrosensitivity”. Part of this loose terminology stems from the lack of a clear molecular understanding of the abundant number of macroscopic observations.4 In a recent review,5 it was pointed out that the origins of many ion-specific effects are to be found in ion binding to the macromolecule6,7 rather than in water structuring effects. A thorough investigation of the possible ion-macromolecular association mechanisms is therefore desirable. So how do ions interact with macromolecules and what is the origin of ion specificity? These are the central questions of this work, and we now give a brief overview of the molecular driving forces that may (or may not) be at play: (1) Coulomb interactions lead to a neutralizing counterion distribution outside charged molecular surfaces.8,9 For highly charged systems, multivalent counterions may even (via ionion correlations) overcompensate for the molecular charge, resulting in a charge reversal.10,11 * To whom correspondence should be addressed. E-mail: mikael.lund@ uochb.cas.cz. (1) Hofmeister, F. Arch. Exp. Pathol. Pharmakol. (Leipzig) 1888, 24, 247260. (2) Kunz, W.; Henle, J.; Ninham, B. Curr. Opin. Colloid Interface Sci. 2004, 9, 19-37. (3) Baldwin, R. L. Biophys. J. 1996, 71, 2056-2063. (4) Kunz, W. Pure Appl. Chem. 2006, 78, 1611-1617. (5) Zhang, Y.; Cremer, P. Curr. Opin. Chem. Biol. 2006, 10, 658-663. (6) Kalra, A.; Tugcu, N.; Cramer, S.; Garde, S. J. Phys. Chem. B 2001, 105, 6380-6386. (7) Shimizu, S.; McLaren, W. M.; Matubayasi, N. J. Chem. Phys. 2006, 124, 234905. (8) Evans, D. F.; Wennerstro¨m, H. The Colloidal Domain: Where Physics, Chemistry, Biology, and Technology Meet; VCH Publishers: New York, 1994. (9) Israelachvili, J. Intermolecular and Surface Forces, 2nd ed.; Academic Press: London, 1991. (10) Guldbrand, L.; Jo¨nsson, B.; Wennerstro¨m, H.; Linse, P. J. Chem. Phys. 1984, 80, 2221. (11) Pianegonda, S.; Barbosa, M. C.; Levin, Y. Eur. Phys. Lett. 2005, 71, 831-837.

(2) Short-range dispersion or London type interactions are effective for all types of atoms, polar or not, and are screened by salt only to a mild extent.12 Since the high-frequency polarizability is roughly proportional to the volume of the electron cloud of a given species, ion-specific effects will arise. A number of workers have pursued this idea by combining dispersion interactions with continuum electrostatics, investigating colloidal as well as biological systems.13-15 (3) Similar to the air/water interface,16,17 poorly solvated surface groups may lead to a hydrophobic-like attraction18 of large, polarizable ions such as iodide and thiocyanate. In contrast, small highly solvated ions will experience a repulsion as described within classical electrostatics by a reaction field or image charge repulsion.19 (4) Ion pairing, a term coined in the first quarter of the 20th century,20 implies that the interaction between a cation and an anion is sufficiently strong to form a temporary neutral species. The tendency to form ion pairs is ion specific,21 and we can expect different charged surface groups and free ions to pair with varying strength. As a recent example, a study of ion binding to a series of proteins22 showed that the segregation of sodium versus potassium to the protein surface is caused mainly by preferential binding to carboxylic groups. The next question to address is how ion binding is coupled with macromolecular assembly. In traditional double-layer theory,8 two charged surfaces start to interact when their counterion distributions overlap. Hence, any modification of these (12) Netz, R. R. Eur. Phys. J. E 2001, 5, 189-205. (13) Bostro¨m, M.; Williams, D. R.; Ninham, B. W. Phys. ReV. Lett. 2001, 87, 168103. (14) Tavares, F.; Bratko, D.; Prausnitz, J. Curr. Opin. Colloid Interface Sci. 2004, 9, 81-86. (15) Bostro¨m, M.; Tavares, F. W.; Finet, S.; Skouri-Panet, F.; Tardieu, A.; Ninham, B. W. Biophys. Chem. 2005, 117, 217-224. (16) Jungwirth, P.; Tobias, D. J. Chem. ReV. 2006, 106, 1259-1281. (17) Huang, D.; Cottin-Bizonne, C.; Ybert, C.; Bocquet, L. Phys. ReV. Lett. 2007, 98, 177801. (18) Chandler, D. Nature 2005, 437, 640-647. (19) Bo¨ttcher, C. Elsevier: Amsterdam, 1973. (20) Bjerrum, N. K. Dan. Vidensk. Selsk. Mat.-Fys. Medd. 1926, 7, 1-48. (21) Jagoda-Cwiklik, B.; Vacha, R.; Lund, M.; Srebro, M.; Jungwirth, P. J. Phys. Chem. B 2007, 111, 14077-14079. (22) Vrbka, L.; Vondra´sek, J.; Jagoda-Cwiklik, B.; Va´cha, R.; Jungwirth, P. Proc. Natl. Acad. Sci. U.S.A. 2006, 103, 15440-15444.

10.1021/la7034104 CCC: $40.75 © 2008 American Chemical Society Published on Web 02/23/2008

3388 Langmuir, Vol. 24, No. 7, 2008

Lund et al. Table 1. Charges and Lennard-Jones Parameters Using the Combination Rules σij ) (σi + σj)/2 and Eij ) (EiEj)1/2

Figure 1. Two models of a macromolecule with a nonpolar core as well as charged groups that can be either solvent exposed (A) or buried beneath the surface (B).

distributions (via salt screening, ion binding, etc.) will modulate the effective intermolecular interaction. For example, iodide and thiocyanate have been found to effectively induce attractive interactions between positively charged proteins,23-26 indicating that the ions accumulate at the surface so as to lower the overall charge or, rather, decrease the range of the neutralizing counterion distribution. Ion binding will further affect the solvation free energy of the macromolecule, and this combined with the intermolecular potential will govern the salting in and out properties. From the above discussion, it is clear that the study of specific ion binding and the ensuing implications for solution stability are highly nontrivial. In this study, we use atomistic Molecular Dynamics (MD) simulations to investigate the microscopic distribution of ions around a single charged macromolecule. Due to the multitude of interactions taking place at the interface of complex molecules, a clear identification of the responsible driving forces may prove difficult. Hence, more direct physical insight can be gained by mimicking the system using a coarse grained representation that encompasses major features such as dispersion, hydrophobic, and ionic contributions. This is accomplished by invoking a model where the macromolecular core is treated as a large, nonpolar sphere with charges distributed on the surface as shown in Figure 1. A similar model has previously been employed for the study of hydrophobic interactions (i.e., solvent averaged effective potentials) between proteinlike nanospheres, albeit with no focus on ion-specific effects.27 A related study focuses on the hydration of confined plates with hydrophobic and hydrophilic patches.28 As for the choice of anions, we investigate the two extremes from the halide group, fluoride and iodide; the former represents a small, highly solvated species, while the latter exemplifies a large, poorly hydrated ion.

Model and Simulation Details The configurational space is sampled in the NpT-ensemble using molecular dynamics (MD) simulations in a cubic box with periodic boundaries. Energies and forces are based on the following pair potential function comprised of Coulomb and Lennard-Jones (LJ) interactions,

Na Cs F I HC Hw Ow

σ/nm

/(kJ‚mol-1)

q/e

0.24 0.38 0.31 0.51 1.6 0 0.32

0.54 0.42 0.84 0.42 0.053 0 0.65

+1 +1 -1 -1 0 0.4238 -0.8476

them, σ and  are their combined LJ parameters, e is the electron unit charge, and 0 is the permittivity of vacuum. Detailed interaction parameters are given in Table 1. Long-ranged electrostatic corrections are accounted for by the particle mesh Ewald summation method,29 and we use a spherical cutoff of 1.35 nm for both electrostatic and Lennard-Jones pair interactions. The diameter of the nanosphere is 1.6 nm, and its LJ parameter, , is chosen such that water oxygens experience a net repulsion of ∼1 kT when penetrating the surface by 1 Å (rij ) σij - 1 Å). Charges are placed either 1 Å above the sphere or 1 Å below the spherical surface as shown in Figure 1. This is to mimic two situations of solvent accessibility of the charged surface groups. The angular distribution of the surface charges is such that their mutual electrostatic energy is at a minimum. For comparison, we also simulate a charged sphere of a smaller size (σ ) 0.56 nm) as well as a noncharged sphere of the original size. All simulations contain a background sodium ion concentration of 160 mM and an evenly mixed number of neutralizing flouride and iodide anions. A typical simulation cell (1 atm, 298 K) contains around 4000 SPC/E water molecules,30 one fixed nanosphere, counterions, and a background salt concentration corresponding to that of physiological conditions. Simulations are equilibrated for 400 ps and sampled for 30-60 ns with a time step of 2 fs, which is sufficient to converge the potential of mean force between ions and the macrosphere to within a few tenths of a kT. All MD results are generated using the GROMACS31 simulation package version 3.3.1. To extract the overall ion binding to the macrosphere, we integrate the radial distribution functions, g(r). This yields approximate thermodynamic binding constants,32

Kion ≈

∫∼0∞ (g(r) - 1) r2 dr

(2)

where we note that long-ranged, attractive interactions will dominate due to the r-squared weighting. The fraction KI/KF is used as a measure of the relative binding of iodide versus fluoride.

Results

where q represents the particle charges, r is the distance between

Interaction with an Uncharged Sphere. Large ions with a low surface charge density (iodide, thiocyanate, etc.) are in general poorly solvated, and due to the large cohesiveness of water these solutes may effectively be expelled to other low solvated regions such as nonpolar molecular interfaces.6 This is indeed the case for iodide as shown in Figure 2 where we observe a small but distinct free energy minimum (g(r) maximum) in the interaction with a neutral nanosphere. Conversely, fluoride being strongly

(23) Lafont, S.; Veesler, S.; Astier, J.; Boistelle, R. J. Cryst. Growth 1997, 173, 132-140. (24) Piazza, R. Curr. Opin. Colloid Interface Sci. 2004, 8, 515-522. (25) Finet, S.; Skouri-Panet, F.; Casselyn, M.; Bonnete, F.; Tardieu, A. Curr. Opin. Colloid Interface Sci. 2004, 9, 112-116. (26) Piazza, R.; Pierno, M. J. Phys.: Condens. Matter 2000, 12, A443-A449. (27) Dzubiella, J.; Hansen, J.-P. J. Chem. Phys. 2004, 121, 5514-5530. (28) Giovambattista, N.; Debenedetti, P.; Rossky, P. J. Phys. Chem. C 2007, 111, 1323-1332.

(29) Essmann, U.; Perera, L.; Berkowitz, M.; Darden, T.; Lee, H.; Pedersen, L. J. Chem. Phys. 1995, 103, 8577-8593. (30) Berendsen, H. J. C.; Grigera, J. R.; Straatsma, T. P. J. Phys. Chem. 1987, 91, 6269-6271. (31) Berendsen, H.; Spoel, D.; Drunen, R. Comput. Phys. Commun. 1995, 91, 43-56. (32) Wennerstro¨m, H. Statistical mechanical description of surfactant selfassmbly. In Organized solutions: surfactants in science and technology; Lindman, B., Friberg, S., Eds.; Marcel Dekker: New York, 1992; p 410.

uij )

e2qiqj + 4ij[(σij/rij)12 - (σij/rij)6] 4π0rij

(1)

Specific Ion Binding to Macromolecules

Langmuir, Vol. 24, No. 7, 2008 3389

Figure 2. Radial distribution of iodide, fluoride, and water outside a neutral LJ sphere of diameter 1.6 nm. Note the nonlinear scale of the right-hand axis, showing the potential of mean force (PMF).

Figure 3. Average Lennard-Jones and Coulomb self-energy of a single iodide ion as a function of its distance from the neutral macrosphere.

solvated is repelled from the nanosphere. We also observe a slight water structuring, typical for relatively hard surfaces; in contrast, no density structuring is seen at the air/water interface16 which may be regarded as “infinitely” soft. To determine the relative importance of dispersion interactions, we calculate the statistical mechanical average of the total Lennard-Jones and Coulomb energy for a single anion, j, as a function of the distance, R, to the macrosphere, N

uj(R) ) 〈

uij(R)〉 ∑ i*j

(3)

This energy decomposition is shown for iodide in Figure 3. As for the LJ energy, noise makes it difficult to decipher the distance dependent behavior, but nonetheless, it is weak and hardly sufficient to account for the 1 kT minimum observed in the potential of mean force. Further, the Coulomb energy is strongly repulsive near the sphere, manifesting the loss of ion-dipole interactions when moving iodide closer to the nonpolar surface. Thus, iodide’s affinity to the surface has the character of a hydrophobic attraction, driven mainly by the solvent.33 Increasing the temperature by 25 K, we observe a decrease in the ionbinding free energy indicative of an entropic repulsion which at the g(r) maximum amounts to

T d ln g(rmax) dG )≈ 4 kT dT dT

-TS ) T

(4)

at 298 K. Similarly, small-angle X-ray scattering experiments of protein-protein interactions25 show evidence of a short-range, ion-specific attraction that is also weakened at elevated temperatures, compatible with the idea that specific ion binding is a central driving force for Hofmeister effects on macromolecular assembly. (33) Garde, S.; Hummer, G.; Garcı´a, A. E.; Paulaitis, M. E.; Pratt, L. R. Phys. ReV. Lett. 1996, 77, 4966-4968.

Figure 4. Radial distribution of fluoride and iodide outside a nanosphere with eight attached charges that are either 1 Å above the surface (top) or 1 Å below the surface (bottom). The inset shows the case where all charges are placed in the center of the sphere.

The molecular solvent simulations performed here suggest that dispersion interactions may contribute less to the ion specificity than hitherto thought and that a major contribution stems rather from solvation effects. It is to be noted that treating the hydrophobic core as a LJ particle somewhat overestimates the distance dependent decay of the dispersion interactions.9 A straightforward improvement would be to construct the nanosphere from a number of smaller spheres, and work is in progress to implement this with proteins at the amino acid level.34 The presented results are obtained from simulations with a nonpolarizable force field. For iodide at the air/water interface, inclusion of atomic polarizabilities results in an enhanced surface affinity16 and a similar effect may present itself at molecular interfaces. Preliminary results obtained by calculations with explicit polarizabilities of both solvent molecules and ions indeed intensify the hydrophobic attraction, yet the qualitative picture is maintained. Charged Groups. We now attach eight explicit unit charges to the spherical core, with each one having the same size as a sodium ion. Let us first examine the situation where the ions are situated 1 Å aboVe the spherical core (see Figure 1, left), corresponding to solvent exposed groups as found in micelles, peptides, and other macromolecules. Figure 4 shows that both halides are attracted to the nanoparticle and that the larger iodide anion binds to the sphere significantly more than fluoride (KI/KF ≈ 2). Furthermore, the iodide peak is wide, whereas fluoride exhibits a more narrow spike. To elucidate the structural differences between iodide and fluoride, we resort to spatial isodensity plots35 of both ions outside the nanoparticle. Figure 5 demonstrates the striking difference between the two halides: fluoride binds highly specifically to the solvent exposed charged groups, while iodide tends to occupy the surface area in between these groups in a more delocalized fashion. The fact that fluoride indeed has a preference for the charged groups is further illustrated (34) Lund, M.; Jo¨nsson, B. Biophys. J. 2003, 85, 2940-2947. (35) Humphrey, W.; Dalke, A.; Schulten, K. J. Mol. Graphics 1996, 14, 3138.

3390 Langmuir, Vol. 24, No. 7, 2008

Lund et al.

Figure 5. Left: spatial isodensity plot of iodide (red, c ) 4.3 × 10-3 1/Å3) and fluoride (green, c ) 2.8 × 10-2 1/Å3) outside the nanosphere with eight solvent exposed positively charged groups (dark blue). Right: Same as on the left, but with half of the (sodiumsized) surface charges inflated to the size of cesium (yellow).

in Figure 4 (bottom), where we have shifted the charges 1 Å below the surface of the nanosphere as sketched in Figure 1. The charges are thus less accessible for the strongly solvated fluoride anion, and we see a decreased binding of this species (KI/KF ≈ 3). By contrast, iodide no longer needs to compete with fluoride for surface sites, and we observe a slight increase in its binding to the macromolecule. In summary, while both halide anions are attracted to the nanosphere, they do so in fundamentally different manners. Fluoride binds with strong, specific charge-charge interactions, while iodide associates via delocalized interactions to the lesser solvated surface between the charged groups due to a combination of hydrophobic and electrostatic effects. Finally, the inset of Figure 4 (top) shows the effect of moving all eight charges to the middle of the macromolecule, which corresponds to the key approximation in classic DerjaguinLandau-Verwey-Overbeek theory.9,36,37 The iodide peak now corresponds to a free energy minimum of ∼6 kT or 2-3 times more than that when the charges are placed outside the sphere, while the fluoride peak vanishes. Thus, while the central charge approximation may be appropriate in the continuum model, it has a larger impact in an explicit solvent description, in accord with other studies.27 Effect of Charge Density. Based on experimental data for a wide range of systems, an empirical rule of “matching water affinities” has been proposed38,39 to explain ion-specific behavior. Note that within the Born solvation model40 this rule translates into “matching ion sizes”. That is, cationic and anionic groups with similar sizes (or surface charge densities) tend to bind to each other. Our findings support this idea in that fluoride binds to the small charged patches (here, the size of sodium), while iodide stays in regions of lower surface potentials. This ionpairing preference is also manifested experimentally in, for example, activity coefficients, γ, for pure alkali fluoride and iodide solutions41,42 where the excess chemical potential difference ex ∆µIfF ) kT ln(γMF/γMI)

(5)

reveals that fluoride is the preferred halide anion for pairing with sodium (compare Figure 6). In contrast, iodide is the most favorable binding partner to larger cations such as rubidium or cesium. Following this experimental rationale, replacing the sodium-sized, charged patches with cesium should lead to a substantial decrease in fluoride binding. As shown on the righthand side of Figure 5, this is indeed the case: exchanging half (36) Derjaguin, B. V.; Landau, L. Acta Phys. Chim. 1941, 14, 633-662. (37) Verwey, E. J. W.; Overbeek, J. T. G. Theory of the Stability of Lyophobic Colloids; Elsevier Publishing Company Inc.: Amsterdam, 1948. (38) Collins, K. D. Methods 2004, 34, 300-311. (39) Collins, K. D. Biophys. Chem. 2006, 119, 271-281. (40) Born, M. Z. Phys. A 1920, 1, 45-48. (41) Robinson, R. A.; Stokes, R. H. Electrolyte Solutions; Butterworths Scientific Publications: London, 1959. (42) Tien, H. J. Phys. Chem. 1963, 67, 532-533.

Figure 6. Experimental excess chemical potential differences for exchanging iodide with fluoride in aqueous alkali halide salt solutions.

Figure 7. Radial distribution of fluoride and iodide outside a smaller (σ ) 5.6 Å) sphere with eight surface exposed charged groups. The graphical inset shows the isodentity plot of iodide (red) and fluoride (green).

of the sodium patches with cesium results in an increased binding of iodide at the expense of fluoride. It is interesting to note that fluoride binding is sustained only at the remaining sodium-sized patches, emphasizing the local nature of the interaction. Note that, for “real” proteins, the cationic surface groups are of course more complex. To further demonstrate the important balance between hydrophobicity and surface charge, we now increase the surface charge density by decreasing the size of the spherical nanocore (σ ) 1.6 f 0.56 nm) so that this becomes significantly more hydrophilic. The positions of the eight sodium-sized charges are also rescaled so that they remain 1 Å above the surface of the nanosphere. As seen in Figure 7, the picture is now reVersed and fluoride binds more strongly than iodide. We see a pronounced fluoride peak at ∼6.5 Å stemming from direct pairing with the charged patches, and as these are now closer together, the distribution function gets broader. Thus, the size and availability of the charged surface groups are expected to play an important role for ion binding and may explain how weakly hydrated ions can also interact via ion-ion interactions as previously suggested.43 In a recent MD study of ion binding to a small, flexible tripeptide,44 it was shown that fluoride and iodide behave in much the same way as presented here. This hints that the proposed physical mechanisms are valid over a range of molecular categories, compatible with the notion that the specific ion binding is primarily determined by local surface properties.5

Conclusions Using explicit solvent molecular dynamics, we have investigated the distribution of fluoride and iodide anions around a spherical macromolecule both with and without discrete charged patches. In the case of an uncharged macromolecule, we find (43) Zhang, Y.; Furyk, S.; Bergbreiter, D. E.; Cremer, P. S. J. Am. Chem. Soc. 2005, 127, 14505-14510. (44) Fedorov, M. V.; Goodman, J. M.; Schumm, S. Phys. Chem. Chem. Phys. 2007, 9, 5423-5435.

Specific Ion Binding to Macromolecules

that fluoride, being strongly hydrated, is repelled while the less solvated iodide is weakly attracted to the surface. After charging up the nanosphere, both anions are attracted, albeit in two distinct manners: (1) Fluoride, as a small ion, binds by specific anioncation interactions that overcome the repulsion with the hydrophobic core. (2) Iodide, as a large ion, binds by ion pairing combined with a more delocalized, hydrophobic attraction. This study suggests that the main contribution to specific ion binding originates from hydrophilic and hydrophobic interactions, while we observe little or no effect from dispersion interactions. Burying the positively charged groups below the surface of the core, iodide binding is strongly favored. If we increase the surface charge density, fluoride dominates the binding. Thus, the specific anion binding is governed by local interactions that may be approximately divided into the two molecular mechanisms proposed above.

Langmuir, Vol. 24, No. 7, 2008 3391

While our generalized coarse grained approach is in qualitative agreement with both bulk electrolyte data and MD simulations of small peptides,41,42,44 the distinct nature of the “real” (bio-)macromolecule is expected to modulate the quantitative binding of small ions. Acknowledgment. For financial support, we thank the European Molecular Biology Organization (EMBO), and for computational resources we thank the LUNARC center at Lund University, Sweden. The center in Prague is supported by the Czech Ministry of Education via Grant LC512. R.V. acknowledges support from the Granting Agency of the Czech Republic (Grant 203/05/H001). The Institute of Organic Chemistry and Biochemistry is supported via Project No. Z40550506. LA7034104